Seamless restarts

[Rulexec]

We need to refactor work with interfaces/raw_db in a way that allow us to start other instance of diffbelt which will take ownership of raw_db without downtime.

After #15 we will see how long our requests can be executed and maybe we need just balancer that can queue incoming requests, wait for processing finish of the old instance and then routing those requests to the fresh instance. But maybe we'll need to do it in a more complex way with internal API where we can close raw_db in old instance, reopen it in the fresh one and then continue to use it from the requests inside old instance via this internal API.